Full Analysis of Palantir CEO’s Critique of OpenAI and Anthropic’s Token-Based Pricing Model and the Rise of Sovereign AI (2026)
In 1517, a monk and theologian nailed 95 theses to the door of the Castle Church in Wittenberg, igniting the fire of the Reformation. Martin Luther was not attacking Christianity. He was attacking the Church’s business model - the sale of indulgences. People were paying for “tokens” of salvation certificates that supposedly reduced punishment in purgatory. In return, the institution accumulated enormous wealth and power, while the true spiritual value for believers remained questionable. Luther called this corruption and a misalignment of incentives: you pay for something that does not deliver proportional benefit, while the institution extracts your most valuable asset - your trust, your fear, and ultimately, your resources.
Five hundred and nine years later, on July 1, 2026, during an interview on CNBC’s Squawk Box, Alex Karp - CEO of Palantir and someone who himself profits from artificial intelligence - did something remarkably similar. He did not attack artificial intelligence. He attacked the dominant business model of today: selling tokens instead of sharing real value, and covertly extracting intellectual property and competitive advantage (“alpha”) from enterprises.
“We’re paying for tokens that create no value,” Karp said. “These people are stealing the weights and alpha of my business.”
The difference today is that this is no longer about purgatory - it is about survival in a competitive economy. It is no longer about the remission of sins, but about the remission of data - your data, your processes, your unique business logic that feeds the models of OpenAI, Anthropic, and other frontier labs, while you are charged by the millions of tokens.
What Exactly Did Alex Karp Say About Token-Based Pricing in AI and the “Theft” of Alpha?
On Wednesday, July 1, 2026, during a regular appearance on CNBC’s Squawk Box discussing Palantir’s new partnership with Nvidia on “Sovereign AI architecture,” Alex Karp did something rare in corporate America: he said out loud what dozens of CEOs have been whispering to him in private for months.
“Something has gone completely wrong,” he stated. “The basic view among enterprises in this country is: ‘I’m going to chillax and waste my time with tokens, I’m going to get no value, and they’re going to get my IP.’”
This was not mere criticism. It was a direct accusation against OpenAI, Anthropic, and the entire frontier AI model ecosystem: they are selling tokens instead of sharing value, while simultaneously extracting the most valuable asset any enterprise possesses - its alpha (competitive advantage), its business processes, its data, and the way all these elements interconnect.
“Theft of Intellectual Labor” - Not a Metaphor, But a Business Model
Karp did not use the word “theft” lightly. He repeated it several times:
“We’re paying for tokens that create no value. These people are stealing the weights and alpha of my business.”
“Weights” here is not merely a technical term. It is a metaphor for how a real business actually operates - how customers are segmented, how decisions are made, and how data from ERP, CRM, supply chain, and operational systems are interconnected. When a company sends prompts to GPT-4o or Claude 4, it is not simply “getting an answer.” It is training the model on its own alpha.
Frontier labs get paid twice: once through the tokens the customer buys, and a second time through improved models that are later sold to the customer’s competitors. This is a classic double-dipping scheme, dressed in the language of “convenience” and “scale.”
Karp posed the question that should be obvious but which the AI industry stubbornly avoids:
“If I can truly make you a billion dollars tomorrow, wouldn’t I say: ‘I’ll make you a billion and I want 30%’? Why are they charging for tokens if it’s really that valuable?”
This is the core of the problem. Token-based pricing is convenient for the provider but completely inadequate for the client when it comes to transformative technology. It turns AI into a consumption commodity rather than a partnership for value creation. It creates a perverse incentive: the more tokens you burn (tokenmaxxing), the better it is for OpenAI and Anthropic - regardless of whether it delivers real returns to you.
The “Wealth Tax” That Punishes Successful Companies
Karp went even further. He called this model a “wealth tax” - a tax that does not help the poor but simply punishes those who are still creating value.
“Every single enterprise I deal with, these people are livid. They’re like, ‘I am paying for tokens that create no value. These people are stealing the weights and alpha of my business, and they’re creating a wealth tax that does not help the poor, it just punishes.’”
This is a deeply political and economic observation. At a time when governments worldwide are discussing how to tax AI profits, AI companies themselves are already imposing a hidden tax on corporate America - in the form of token costs plus the loss of intellectual property.
Palantir and the Sovereign AI Model
Karp did not stop at criticism. He offered an alternative - and it is exactly what Palantir has been selling for years: sovereignty in artificial intelligence (Sovereign AI).
Instead of sending your data to OpenAI’s cloud, Palantir (in partnership with Nvidia) offers an architecture in which:
- Data never leaves the controlled environment;
- The model is not retrained on your alpha;
- “Ontology” (Palantir’s application layer) turns the large language model into a safe, precise, and integrated tool for real business processes;
- The client retains control over compute, models, data stack, and most importantly - its alpha.
“Technical customers want control over their compute, their models, their data stack, and their alpha. They want to know they own the means of production. It’s not being transferred to someone else.”
This is not merely a technological difference. It is a philosophical and economic one. In a world where AI is becoming infrastructure, the question of “who owns the means of production” has once again become central - just as it was in the 19th and 20th centuries.
Table 1. Comparative Analysis of Two Models for Enterprise AI Deployment
| Aspect | Token-Based Model (OpenAI, Anthropic) | Sovereign Model (Palantir + Nvidia) |
|---|---|---|
| Pricing | Per number of tokens processed | Outcome-based, control-oriented, value-linked |
| Data Control | Data leaves the enterprise | Data remains in sovereign environment |
| IP Risk | High | Minimal |
| Application Layer | Weak or absent | Strong ontology layer |
| Ownership of Means of Production | Transferred to provider | Remains with the client |
What Does Alex Karp’s Critique Mean for Capital Markets and AI Investments?
1. Palantir (PLTR) gains a powerful narrative tailwind. The company has been talking about “ontology” and “sovereignty” for years while the market was obsessed with raw model scaling. Now enterprise and government clients are beginning to hear the same message Karp has been delivering. It is no coincidence that the stock reacted positively after the interview.
2. Frontier labs (OpenAI, Anthropic, Google DeepMind) face increasing pressure to evolve their business models. Tokens work well for consumer and startup segments that have little to lose. For Fortune 500 companies and critical infrastructure, they are becoming less acceptable. Expect more hybrid or on-premise offerings, greater transparency on data retention, and possibly outcome-based pricing experiments.
3. Nvidia benefits from both sides - chip sales for cloud training and partnerships for sovereign/on-premise stacks.
4. The broader AI investment thesis is shifting. “Who has the best model?” is gradually giving way to “Who controls the data, integration, and risk?” This favors companies with deep domain expertise and strong enterprise relationships over pure model plays.
Table 2. Historical Parallels: Public Figures Who Exposed Misalignments in Dominant Business Models
| Figure / Work | Year | What They Exposed | Parallel to Karp’s Critique |
|---|---|---|---|
| Martin Luther | 1517 | Sale of indulgences (“tokens of salvation”) | Paying for tokens without real value |
| Warren Buffett | 2002 | Derivatives as “financial weapons of mass destruction” | Misalignment between promised and actual value |
| John Bogle | 1976+ | High fees in the mutual fund industry | Tokenmaxxing without proportional returns |
| Upton Sinclair | 1906 | Exploitation in the meatpacking industry under the guise of “progress” | “Innovation” that masks systemic value extraction |
| Jaron Lanier | 2010s | Use of human data and creativity without compensation | Theft of intellectual labor and alpha |
The Voice of Pragmatic America in the AI Industry
Alex Karp is not a typical Silicon Valley CEO. He does not sell utopia. He sells control and results - especially to clients for whom mistakes are extremely expensive (defense, healthcare, critical infrastructure, large corporations).
When he says “something has gone completely wrong,” it is not mere marketing. It is a signal that the most serious AI customers are no longer willing to pay for hype. They want value, security, and ownership.
The AI revolution is not over. But its first phase - the phase of “give us all your data and we’ll sell you tokens” - is beginning to meet serious resistance. Alex Karp has simply articulated it most clearly and loudly.
And capital markets are starting to listen.
The Beginning of a Necessary Reformation in AI
Martin Luther did not want to destroy the Church. He wanted to return it to its true mission - or at least that is what he claimed. History showed, however, that when a dominant institution begins charging for “tokens of salvation” without delivering proportional value while accumulating power and information, a reaction becomes inevitable.
Today we are witnessing a similar moment in artificial intelligence. Alex Karp is not Luther and Palantir is not a new Protestant church. But his voice - the voice of an insider who profits from the system yet refuses to stay silent about its deformations - marks the beginning of a necessary Reformation in how artificial intelligence is built, priced, and controlled.
The question is no longer just “how many tokens can we sell?” The question is: who owns the means of production in the age of artificial intelligence - the data, the alpha, the models, and the ability to extract value from them?
History does not forgive institutions that charge for promises instead of results for too long. The AI industry still has a chance to avoid the fate of the late medieval Church. But if it continues selling tokens while stealing alpha, someone else - whether Karp, regulators, or the enterprises themselves -will nail their own 95 theses to the digital door.
For investors, for executives, and for anyone who believes AI should amplify rather than undermine the competitive advantage of its users, the moment of choice has arrived.
Frequently Asked Questions (FAQ)
1. What is the precise historical parallel between Martin Luther and Alex Karp?
Martin Luther exposed the sale of indulgences - literal “tokens” for which people paid without receiving proportional spiritual value, while the institution extracted wealth and power. Alex Karp is doing the same today: enterprises pay for AI tokens that often deliver no real business value, while frontier labs extract their alpha and intellectual property. In both cases, it is a public exposure of misalignment between price and actual benefit.
2. What exactly does “alpha” and “weights” mean in this context?
“Alpha” refers to an enterprise’s unique competitive advantage - its data, processes, customer relationships, decision-making logic, and the way all these elements interconnect. “Weights” is a metaphor for the importance and interconnections within that data. When a company uses OpenAI or Anthropic APIs, it transfers part of this alpha into the provider’s models, which can later be used for other clients.
3. Why is token-based pricing so problematic for large enterprises?
It creates a perverse incentive: the more tokens are consumed, the more the provider earns - regardless of whether the customer receives real value. When dealing with transformative technology, this makes little sense. If AI can truly generate hundreds of millions or billions in value, it would be logical to share a portion of that value (outcome-based pricing) rather than simply charging for usage. Currently, enterprises pay for the process, not the result, while simultaneously losing control over their most valuable asset
4. What is Palantir’s “ontology” and why is it so important for sovereignty?
Ontology is the application layer that turns a raw language model into a safe and useful tool for a specific enterprise. It defines what data means to the business, how elements connect, and what rules apply. Because of it, the model does not touch raw data, does not cache it, and does not use it to train other models. This is the key mechanism that allows the client to retain full control over data, alpha, and the means of production.
5. What are the investment implications of this critique for Palantir and frontier AI labs?
For Palantir, the critique strengthens the “Sovereign AI” narrative and positions the company as a defender of enterprise control - a factor supporting long-term growth with government and large corporate clients. For OpenAI, Anthropic, and similar companies, it creates pressure to develop hybrid models, greater data transparency, and possibly new pricing structures. The broader industry trend is shifting from “who has the biggest model” to “who controls the data and risk” - a shift that favors companies with strong application layers and enterprise trust.
Important notice
This article is for information and education only. It is not financial advice, an investment recommendation or professional advice of any kind. The views expressed are the author’s personal opinions and are not recommendations to buy, sell or hold securities, assets or investments. Investing involves substantial risk, including the possible loss of all capital. Past performance is not a guarantee of future results. Each reader should conduct independent research, consult a qualified financial adviser and weigh financial circumstances, risk tolerance and investment objectives before taking any decision. The author and borsite.net accept no liability for any loss arising from the use of the information in this article.